Episodic Memory in Agentic Frameworks: Suggesting Next Tasks

TL;DR

This paper introduces an episodic memory system for agentic frameworks with LLMs, enabling better next-task suggestions by leveraging past workflows, reducing hallucination, and avoiding extensive fine-tuning.

Contribution

It proposes a novel episodic memory architecture that enhances task recommendation in agentic workflows without relying solely on LLMs or extensive fine-tuning.

Findings

Improved next-task suggestion accuracy.

Reduced reliance on hallucination-prone LLM outputs.

Effective retrieval of relevant past workflows.

Abstract

Agentic frameworks powered by Large Language Models (LLMs) can be useful tools in scientific workflows by enabling human-AI co-creation. A key challenge is recommending the next steps during workflow creation without relying solely on LLMs, which risk hallucination and require fine-tuning with scarce proprietary data. We propose an episodic memory architecture that stores and retrieves past workflows to guide agents in suggesting plausible next tasks. By matching current workflows with historical sequences, agents can recommend steps based on prior patterns.